fix(gateway): track background watcher tasks in _background_tasks

[sprmn24]

What does this PR do?

Four asyncio.create_task() calls in gateway/run.py were launched fire-and-forget — the task handles were discarded immediately, so _background_tasks never tracked them. On shutdown, these tasks could not be cancelled or joined, causing orphaned tasks and silently dropped exceptions.

Type of Change

• Bug fix

Root Cause

The existing pattern used throughout the file is:

task = asyncio.create_task(...)
self._background_tasks.add(task)
task.add_done_callback(self._background_tasks.discard)

Three watcher tasks and the signal handler's runner.stop() call skipped this registration entirely.

Changes Made

• _run_process_watcher (recovered watchers loop) — each task now stored and registered with discard callback
• _session_expiry_watcher — task handle stored and registered in _background_tasks
• _platform_reconnect_watcher — task handle stored and registered in _background_tasks
• signal handler runner.stop() — stored in _stop_task with exception-logging done callback (_background_tasks not accessible in signal handler scope)

How to Test

1. Start gateway with a platform that fails at startup (triggers reconnect watcher)
2. Send Ctrl+C during active processing
3. Observe clean shutdown with no Task was destroyed but it is pending warnings in logs

Checklist

• Follows existing _background_tasks pattern used throughout the file
• No new dependencies
• Only touches the affected task creation sites

[alt-glitch]

Likely duplicate of #6790 — same fix: register fire-and-forget watcher tasks in _background_tasks set in GatewayRunner.

[alt-glitch]

Likely duplicate of #6790 — same fix: register fire-and-forget watcher tasks in _background_tasks set in GatewayRunner.

[sprmn24]

Likely duplicate of #6790 — same fix: register fire-and-forget watcher tasks in _background_tasks set in GatewayRunner.

Thanks for the heads up. Closing - #6790 covers the same fix and goes further with a dedicated helper and regression test.